HAL Id: hal-02933278
https://hal.univ-angers.fr/hal-02933278
Submitted on 8 Sep 2020
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Genetic determinism of prickles in rose
N. Zhou, K. Tang, J. Jeauffre, T. Thouroude, D Lopez Arias, F. Foucher, L.
Hibrand-Saint Oyant
To cite this version:
N. Zhou, K. Tang, J. Jeauffre, T. Thouroude, D Lopez Arias, et al.. Genetic determinism of prickles in rose. TAG Theoretical and Applied Genetics, Springer Verlag, 2020, 19 p. �10.1007/s00122-020- 03652-7�. �hal-02933278�
1
Genetic determinism of prickles in rose
1
NN Zhou1,2, KX Tang2, J. Jeauffre1, T. Thouroude1, D. Lopez Arias1, F. Foucher1*& L.
2
Hibrand-Saint Oyant1*.
3
1 Université d’Angers, Agrocampus-Ouest, INRAE, GDO-IRHS (Genetics and Diversity of Ornamental Plants,
4
Institut de Recherche en Horticulture et Semences), SFR 4207 QUASAV, 49071 Angers, France
5
2 National Engineering Research Center for Ornamental Horticulture; Flower Research Institute, Yunnan Academy
6
of Agricultural Sciences, Kunming 650231, China
7
* both authors contributed equally to the work.
8
Corresponding author: zhouning1116@aliyun.com
9
Key message: The genetic determinism of prickle in rose is complex,
10
with a major locus on LG3 that controls the absence/presence of
11
prickles on the rose stem.
12
Declarations
13
Author Contributions 14
Contributions 15
NN Zhou designed the projects and obtained funding, planned and performed 16
experiments, collected and analyzed data, and drafted the manuscript. F. Foucher and 17
L. Hibrand-Saint Oyant were responsible for piloting and supervising the project, and 18
for revising the manuscript. KX Tang contributed to designing the project and obtaining 19
funding. J. Jeauffre led the qPCR experiment and analyzed data. T. Thouroude prepared 20
the plant cuttings of F1 individuals in the greenhouse. D. Lopez Arias modified the 21
genetic map and helped to develop the R/qtl script.
22
Orcid-ID 23
Zhou NN: orcid.org/0000-0002-5208-6788 24
Tang KX: orcid.org/0000-0003-3807-784X 25
Jeauffre J: orcid.org/0000-0001-6770-0552 26
Thouroude T: orcid.org/0000-0001-7908-7353 27
Lopez Arias D: orcid.org/0000-0001-8129-2786 28
2 Foucher F: orcid.org/0000-0002-3693-7183
29
Hibrand-Saint Oyant L: orcid.org/0000-0002-4451-8798 30
Abstract
31
Rose is one of the major ornamental plants. The selection of glabrous cultivars is an 32
important breeding target but remains a difficult task due to our limited genetic 33
knowledge. Our objective was to understand the genetic and molecular determinism of 34
prickles. Using a segregating diploid rose F1 population, we detected two types of 35
prickles (glandular and non-glandular) in the progeny. We scored the number of non- 36
glandular prickles on the floral and main stems for three years. We performed QTL 37
analysis and detected four prickle loci on LG1, 3, 4 and 6. We determined the credible 38
interval on the reference genome. The QTL on LG3 is a major locus that controls the 39
presence of prickles, and three QTLs (LG3, 4 and 1) may be responsible for prickle 40
density. We further revealed that glabrous hybrids are caused by the combination of the 41
two recessive alleles from both parents. In order to test if rose prickles could originate 42
from a ‘trichome-like structure’, we used a candidate approach to characterize rose gene 43
homologues known in Arabidopsis, involved in trichome initiation. Four of these 44
homologues were located within the overlapping credible interval of the detected QTLs.
45
Transcript accumulation analysis weakly supports the involvement of trichome 46
homologous genes, in the molecular control of prickle initiation. Our studies provide 47
strong evidence for a complex genetic determinism of stem prickle and could help to 48
establish guidelines for glabrous rose breeding. New insights into the relationship 49
between prickles and trichomes constitute valuable information for reverse genetic 50
3 research on prickles.
51
Keywords
52
Trichome, QTL, ZFP5, GIS2, MYB61, MYC1 53
Introduction
54
Rose is the major ornamental plant worldwide with a wide diversity, diverse application 55
forms and an extensive cultivated area. Roses are sold as cut flowers, garden plants, in 56
pots, for essential oil, flower tea and culinary purposes. In past centuries, with the 57
continuous efforts of breeders, more than 33,000 varieties of roses were created (Young 58
2007). However, most of these varieties have persistent prickles on the stem. Prickles 59
can protect against herbivores by deterring them from eating the stem (Ronel and Lev- 60
Yadun 2012; Burns 2014). Furthermore, prickles can be desirable in roses when they 61
are used in hedges to protect properties (as was the case in Reunion Island during the 62
19th century). However, garden roses without prickles are often desirable. Cut roses 63
with prickles are more difficult to handle, harvest and transport and also constitute 64
safety hazards for consumers and workers. Retailers commonly remove prickles from 65
stems prior to sale. Removing the prickles increases labor costs and causes mechanical 66
damage to the stems, which affects vase life and ornamental value. Although a strong 67
market demand to develop roses without prickles exists (Nobbs 1984; Debener 1999;
68
Canli 2003; Canli and Skirvin 2003; Canli and Kazaz 2009), relatively little is known 69
about the genetic and molecular bases of prickle initiation and development.
70
In plants, prickles are described as outgrowths of the epidermis and subjacent layers 71
4 that lack vasculature, and mainly consist of lignin, suberin, cellulose and hemicellulose 72
(Asano et al. 2008; Li et al. 2012). In rose and raspberry, it was thought that prickles 73
were modified glandular trichomes that differentiate at the time of lignification into 74
their final prickle morphologies (Kellogg et al. 2011).
75
Until recently, only a few studies had been published about the molecular regulation of 76
prickle development, but great progress has been made in trichome initiation and 77
development, especially in Arabidopsis. Several transcription factors (TFs) such as 78
MYB, bHLH, WD40, WRKY and C2H2 zinc finger families’ proteins have been 79
identified as being involved in trichome initiation and development (reviewed in 80
Balkunde et al. (2010), Pattanaik et al. (2014), Ma et al. (2016a), Huchelmann et al.
81
(2017) and Chopra et al. (2019)). A trimeric activator complex consisting of MYB 82
(GLABRA1) - bHLH (GLABROUS3/ENHANCER OF GL3) - WDR 83
(TRANSPARENT TESTA GL1) plays a key role in trichome development (Zhang 2003;
84
Kirik et al. 2005; Patra et al. 2013). This trimeric complex finely regulates the temporal 85
and spatial expression of GLABRA2 (GL2) and TRANSPARENT TESTA GL2 (TTG2), 86
determining the fate and pattern of trichome precursor cells (Rerie et al. 1994; Ishida et 87
al. 2008). The bHLH family genes, MYC1 and TT8, belong to the same clade as GL3.
88
AtMYC1 acts as a positive regulator of trichome initiation (Symonds et al. 2011; Zhao 89
et al. 2012), and AtTT8 controls trichome development on leaf margins (Maes et al.
90
2008). AaMYB1 and its orthologue AtMYB61, belonging to the R2R3MYB subfamily, 91
were thought to affect terpene metabolism and trichome development in A. annua and 92
A. thaliana, respectively (Matías-Hernández et al. 2017).
93
5 The active TTG1 trimeric complex can be repressed by R3 MYB subfamily genes:
94
TRY/CPC/TCL1 act as negative regulars by competing with GL1 for binding to GL3 95
(Wang et al. 2008; Wester et al. 2009; Wang and Chen 2014). The active TTG1 96
complex, in interaction with TTG2, regulates the expression of the R3 MYB inhibitors 97
that move to the neighboring cells where they repress trichome initiation (Pesch and 98
Hülskamp 2004; Pesch et al. 2014).
99
Different growth regulators positively affect trichome initiation, such as GA3, cytokinin 100
and jasmonic acid (Traw and Bergelson 2003), through the activation of GL1 (Gan et 101
al. 2006). Different C2H2 zinc-finger proteins such as GLABROUS 102
INFLORESCENCE STEM (GIS), GIS2, GIS3, ZINC FINGER PROTEIN5, 6 and 8 103
(Gan et al. 2006; Gan et al. 2007) include GA and cytokinin signaling pathways (Zhou 104
et al. 2013). The novel transcription factor TRP interacts with ZFP5 and negatively 105
regulates trichome initiation through the gibberellic acid pathway (Kim et al. 2018).
106
In diploid rose, the presence of prickles on the stem was assumed to be controlled by a 107
single dominant gene (Debener 1999; Shupert et al. 2007) located on linkage group 3, 108
LG3 (Linde et al. 2006). Furthermore, two QTLs were detected on LG3 with the scoring 109
of prickle density (Crespel et al. 2002). Using two F1 progenies, Hibrand-Saint Oyant 110
et al. (2018) also identified a large QTL (or two neighboring QTLs) on LG3 (between 111
position 31 Mb - 46.5 MB corresponding to the end of the chromosome 3) and a 112
significant association between position 31 and 32.4 Mb using a GWAS approach. In 113
tetraploid roses, three QTLs were identified in relation to the number of prickles on the 114
stem: two located on LG2 and one on LG3 (Koning-Boucoiran et al. 2009). Using the 115
6 same K5 population with the same phenotype data but a new genetic map, different 116
QTLs were detected on LG3, 4 and 6 and on LG2 (one year) (Bourke et al. 2018).
117
Recently, a WRKY transcription factor, homologous to Arabidopsis TTG2, was located 118
close to a QTL controlling prickle density, and the gene transcripts are differentially 119
accumulated between prickle and prickless roses (Hibrand-Saint Oyant et al. 2018).
120
In this study, our objectives were to decipher the genetic determinism of stem prickles 121
in rose and to characterize candidate genes involved in prickle initiation and 122
development. First, we defined the different types of prickles on the stem and studied 123
them separately. Using an F1 progeny, we detected QTLs and their position in the rose 124
genome sequence. We further analyzed how the alleles of the major QTLs affect the 125
presence of prickles. We identified putative candidate genes (homologues of genes 126
involved in trichome initiation and development in Arabidopsis) and studied their 127
transcript accumulation. That study suggested that prickles and trichomes may carry 128
two different genetic pathways, providing new insights into the relationship between 129
prickles and trichomes.
130
Materials and methods
131
Plant material 132
A progeny of 151 diploid F1 hybrids obtained from a cross between the female Rosa 133
chinensis ‘Old Blush’ (OB) × the male R. x wichurana (RW) was used for map 134
construction (described in Hibrand-Saint Oyant et al. 2018) and QTL analysis. The 135
plants were grown in a field and managed by the Horticulture Experimental Unit 136
(INRAE, Angers, France). The plants were pruned each December. In the following 137
7 spring, new stems developed from the axillary buds from the old pruned stems, and are 138
referred to as “floral stems” since they develop flowers. Later, new stems arise from the 139
base of the plant and are referred to as “main stems”. They remain vegetative in once- 140
flowering individuals and may become floral in continuous-flowering individuals.
141
Phenotypic data collection and analyses 142
To score prickle density, we selected three independent floral and main stems for each 143
F1 progeny and the two parents. The prickle numbers were counted for each selected 144
stem on four internodes (located in the middle of the stem) for three years (2016, 2017 145
and 2018).
146
Statistical analysis and visualization were performed using R version 3.2.3. We 147
visualized the frequency distribution and Q-Q plot using the ‘hist’, ‘legend’, ‘qqnorm’
148
and ‘qqline’ functions. We performed mixed-factorial ANOVA analysis with ‘aov’. A 149
‘shapiro.test’ was used to test the normality of the original data and the ANOVA 150
residuals. When the null hypothesis was negated, ‘kruskal.test’ was used to test if there 151
was any significant difference between the replicate shoots, years and the type of stem 152
variance. ‘pairwise.wilcox.test’ with ‘p.adjust.method =BH’ was used to calculate 153
pairwise comparisons between group levels with corrections for multiple testing. We 154
displayed the distribution of prickle density with a boxplot to compare the difference 155
between the variance using the ggplot2 and ggpubr packages.
156
8 Variance components were estimated with the restricted maximum likelihood (REML) 157
method using the ‘sommer’ package. Phenotype variance components of prickle density 158
were obtained using the following model:
159
Pijlr =µ+ Gi + Sl + Y(l)j + GSil +GYij + ɛijlr, 160
where Pijlr is the phenotypic value of a trait counted on a triplicate stem r of the stem 161
type l of the individual i in the year j, µ is the overall mean, Gi is the random effect of 162
genotype i, Sl is the random effect of stem type l, Y(l)j is the random effect of year j 163
nested in stem type l, GSil is the random interaction between genotype i and stem type 164
l, GYij is the random interaction between genotype i and year j, and ɛijlr is the random 165
residual error.
166
The phenotypic variance (𝜎𝑃2) of stem prickles was divided into the variance of 167
genotypic effect (𝜎𝐺2), genotype × year interaction (𝜎𝐺𝑌2 ), genotype × stem type 168
interaction (𝜎𝐺𝑆2 ), and the residual error variance (𝜎𝐸2).
169
Narrow-sense heritability (h2) was calculated as follows:
170
ℎ2 = 𝜎𝐺2/(𝜎𝐺2+𝜎𝐺𝑌2 /y+𝜎𝐺𝑆2 /s+𝜎𝐸2/𝑦𝑠𝑟) 171
where y is the number of years, r is the number of replication shoots per individual, and 172
s is the number of stem types (PF and PM).
173
Genotypic data 174
The genetic determinism was conducted using the genetic map previously obtained by 175
Hibrand-Saint Oyant et al. (2018) and modified by Lopez-Arias et al. (in prep).
176
QTL Analysis 177
In this study,we performed QTL detection for prickles on the floral (PF) and main (PM) 178
9 stems from data scored in 2016, 2017, 2018 (referred to as PF2016, PF2017, PF2018, 179
PM2016, PM2017 and PM2018, respectively). QTL analyses were carried out using the 180
R/qtl in R version 3.2.3. Based on the non-normal phenotype distribution data, single 181
QTL analysis and LOD scores were calculated using the ‘scanone’ function with non- 182
parametric model (model="np", ties.random = FALSE, method = "em") and the two- 183
part model (model="2part", upper = FALSE) (Boyartchuk et al. 2001).
184
In the non-parametric model, the genome-wide and chromosome-wide significance 185
thresholds of LOD scores were estimated by permutations tests (n.perm = 1000, 186
n.cluster = 20). The Bayesian credible interval was computed with 0.95 and 0.99 187
coverage probabilities. When QTLs for different traits had overlapping 0.95 credible 188
intervals, they were declared to be a potentially “common QTL (cQTL)” (Kawamura 189
et al. 2011). The percent of variance explained by each QTL was calculated by ‘makeqtl’
190
and ‘fitqtl’ with a ‘normal’ model.
191
In the two-part model, the phenotype was separated into two parts: first, the trait value 192
was considered as without (0) or with (1) prickles; if it had prickles, the trait value 193
above zero was assumed to be normally distributed. Three LOD scores for each 194
genomic position were calculated: LOD(p) and LOD(µ) were calculated for binary 195
traits (0 or 1) and non-zero phenotype quantitative traits (> 0), respectively; LOD(p,µ) 196
is simply the sum of the LOD scores from the two separate analyses (Broman 2003).
197
The genome-wide significance thresholds of threeLOD hypotheses were also estimated 198
by 1000 permutation tests and summarized by a 0.05 alpha threshold. The percent of 199
variance explained was calculated by ‘makeqtl’ and ‘fitqtl’ with ‘binary’ and 'normal' 200
10 models for binary(p) and quantitative(µ) traits.
201
Selection of rose candidate genes involved in prickle density 202
Proteins involved in trichome initiation and development were selected in A. thaliana 203
from the TAIR database (https://www.arabidopsis.org) with searching terms GL1, 204
MYB82, MYB61, CPC, TRY, GL3, TT8, MYC1, TTG1, TTG2, ZFP5, ZFP1, GIS2, 205
GIS3, GL2. Rose homologues were searched using BLASTp in the Rosa chinensis 206
Genome v1.0 (Hibrand-Saint Oyant et al. 2018). In addition, we also searched the 207
transcription factors (TF) belonging to the bHLH, WD40, R2R3MYB, C2H2 and 208
WRKY families in rose and which were located on the major cQTL interval of LG3.
209
Using Geneious 9.1.7, ‘Multiple Align’ was performed for the family gene sequences.
210
Conserved domains were used to build phylogenetic trees using the ‘Geneious Tree 211
Builder’ tool with the Jukes-Cantor genetic distance model and the UPGMA tree build 212
method. The rose candidate genes were named according to the following nomenclature 213
corresponding to Rc (for Rosa chinensis) added to the corresponding gene name in 214
Arabidopsis, e.g., RcTTG2 for the rose TTG2 homologue.
215
Gene expression analysis 216
Primers were designed using Primer Premier 5.0 software. To ensure the specificity of 217
the primers, forward and reverse primers were designed in the last exon and in the 218
beginning (first 100 bp) of the 3'UTR. Primer length was between 18 and 25 bp, product 219
length was between 70 and 200 bp, GC content was between 40% and 60%, and the 220
annealing temperature was 58~65 ℃. Primers are listed in Supplementary Table 1.
221
For the qPCR experimental design, we selected four contrasting once-flowering 222
11 individuals from the OW progeny for prickle density: two with no prickles (OW9067 223
and OW9068) and two with prickles (OW9137 and OW9071 with means of 2.5 and 4 224
prickles per internode on the main stem, respectively). The materials were sampled in 225
April 2018 in a greenhouse (three biological replicates). Stems were harvested at 226
different stages of prickle development for roses with prickles, and stems at the same 227
stages for roses without prickles (Supplementary Figure 1). Total RNA was extracted 228
using the NucleoSpin RNA Plus-XS kit for early stages (I and IIa) and using the 229
NucleoSpin RNA Plus-kit for later stages (IIb, IIc and III) according to the 230
manufacturer’s instructions, with minus modifications (2%PVP40 in lysis buffet). The 231
purity of the RNA was checked on 1% agarose gel, and the concentration was measured 232
by an UV spectrophotometer. cDNA was obtained from 500 ng of total RNA using 233
iScriptTM Reverse Transcription Supermix for RT-qPCR (Bio-Rad, Hercules) 234
accordant to the manufacturer’s instructions.The purity and quality of the cDNA were 235
checked by performing PCR amplification with a blank and RW’s DNA sample control, 236
and the concentration was measured with a UV spectrophotometer. RT-qPCR reactions 237
were performed using the soAdvancedTM Universal SYBR® Green Supermix (Bio- 238
Rad) on the CFX Connect Real-time PCR system (Bio-Rad). The gene efficiency was 239
evaluated with a serial dilution of the thirty cDNAs pooling (1:10, 25, 50, 100, 250, 240
500, 1000). A 1:25 dilution of each cDNA was used to analyse the expression pattern 241
of ten candidate genes and two reference genes UBC and TCTP (Randoux et al. 2012).
242
Data collection was performed using the Bio-Rad CFX Maestro1.1. Amplification 243
efficiency of the ten genes ranged from 90.5-104.1%. The reference genes UBC and 244
12 TCTP presented high expression stability in all the samples.
245
For the technical replicates, potential outliers were excluded from the analysis when the 246
standard deviation (SD) of samples is higher than 0.5. Only seven technical replicates 247
(seven out of 390) were excluded: CPC in PIIb (biological group A, C) and in NPIIc 248
(group A), GIS2 in NPIIc (group C), NPIII (group B) and PIII (group B).
249
Normalized expression (∆∆Cq) was calculated using Bio-Rad Maestro1.1 software by 250
applying the ‘gene study’ tool. The cluster analysis for sample and target genes with the 251
mean value of normalized expression was performed using R software with the 252
‘pheatmap’ package. NP samples were used as controls to compare the normalized 253
expression of genes between P and NP samples in the different stages. |Fold change 254
(FC)| >2 and the Wilcoxon signed rank test (p-value < 0.05) as cut-off values in scatter 255
plots were used to demonstrate the significant difference of normalized expression 256
between P and NP samples. NPI was used as a control to visualize the relative 257
normalized expression during stem development in prickle and glabrous stems.
258
Results
259
1-Type, distribution and genetic variability of stem prickles in OW progeny 260
Both parents of the F1 progeny (‘Old Blush’ and R. x wichurana) present prickles on 261
their stems (Figure 1a) (a mean of around ten prickles on four internodes). In the F1
262
progeny, hybrids without prickle can be observed (14 out of 151; no prickles on the 263
three stems scored over three years). These hybrids with glabrous stems (Figure 1b) 264
are referred to as ‘prickless’ individuals (Figure 1c). Out of the 137 F1 individuals with 265
prickles (Figure 1b), nine hybrids were nearly prickless (prickle number on four 266
13 internodes < 1 for three scored years and two types of stems; Figure 1d, Supplementary 267
Figure 2), and numerous stems were glabrous for some individuals, whereas other stems 268
presented a few prickles (variable between the genotypes with unstable states between 269
years and types of stems). Macroscopic analysis shows that parents that present prickles 270
originated from a ‘non-glandular’ structure. These prickles are referred to as Non- 271
Glandular Prickles (NGP). All the F1 prickly individuals (137 out of 151) have NGP.
272
However, some individuals with NGP prickles also present another type of prickle (27 273
out of 137). These prickles present a ‘glandular head’ structure and are referred to as 274
Glandular Prickles (GP) (Figure 1b and 1c, Supplementary Figure 2). Since the 275
presence of GP in the OW progeny is rare (27 and 12 out of 151 on flowers and main 276
stems, respectively; Figure 1d) and very irregular, we decided to consider only NGPs 277
in this study.
278
For the 151 F1 progenies, the number of NGPs on four internodes of floral (PF) and 279
main (PM) stems ranged from 0 to 52 and from 0 to 48, respectively (Supplementary 280
Figure 2). Among them, OW9106 and OW9107 have a much higher prickle density (28 281
to 52 in the scored years) than the others. As for the two parents, OB and RW have an 282
average of 11.1 and 8.7 on PF, and an average of 11.8 and 9.2 on PM, respectively 283
(Figure 1a, Table 1). The ranges of NGP numbers in the F1 hybrids were obviously 284
beyond the values of the two parents, indicating a transgressive segregation.
285
The Shapiro-Wilk normality test and the Q-Q plot of original data (W = (0.692~0.936), 286
p < 2.96 ×10-8) (Supplementary Figure 2) and variance residuals (W = 0.88591, p- 287
value < 2.2e-16) showed that the NGP densities on stems in the F1 population were not 288
14 normally distributed. We tried to transform data (log10, SQRT, box-cox) to make them 289
normal but without success. The Kruskal-Wallis test reveals a genotype effect, a year 290
effect and a stem effect (Table 1). The high heritability (h2 ≈ 0.97) demonstrated that 291
the genetic analyses of stem prickle of this population were reliable (Table 1).
292
2- QTL analysis 293
2.1 Non-parametric QTL analysis 294
For the female and male maps, strong QTLs were detected on LG3 for the two types of 295
stems and for the three years (Figure 2 and Table 2). The LOD scores are higher for the 296
male map (between 8 and 11.5) and relatively lower for the female map (between 2.3 297
and 6.2). These QTLs explained between 6.65 to 37.4% of the phenotypic variance. The 298
locations of these QTLs are very close. Indeed, on the female map, the marker at the 299
peak of the QTLs is the same for both types of stems (Rh12GR_16570_782, 51.1 cM, 300
located on the chr3 at 44,459,262 bp according to the Rosa chinensis Genome v1.0 301
(Hibrand-Saint Oyant et al. 2018)), except for PM2018 (Rh12GR_34665_95, 45.7 cM, 302
located on chr3 at 41,401,120 bp). On the male map, for the two types of stems and for 303
the three years, the marker with the highest LOD for the QTLs detected on LG3 is the 304
same, Rh12GR_52506_1218 (42.6 cM on the LG3, 42,317,122 bp on Chr3), which is 305
the terminal marker on the genetic map but not on the physical map.
306
Furthermore, if we consider the common 0.95 Bayesian credible interval of these QTLs 307
on LG3 on the female and male maps, all intervals are overlapping (Table 2 and Figure 308
4). For the female map, the interval on LG3 was 40.38-53.75 cM, which corresponds 309
to the interval 36,517,224-46,440,369 bp on the physical map of chr3 (Figure 4a), and 310
15 for the male map, the interval on LG3 was 37.69-42.55 cM, corresponding to the 311
interval 41,648,024-42,317,122 bp on the physical map (Table 2, Figure 4b).
312
On LG4, QTLs were only detected on the female map for the main stem for the three 313
years (Figure 2, Table 2). The peak marker Rh12GR_60129_183 located at 30.6 cM, 314
which is located on chr4 at 52,239,028 bp, explained 10.35 to 13.18% of the observed 315
variance depending on the year of the phenotypic variance in the single QTL model.
316
The common 0.95 credible interval on LG4 was 20.53-48.59 cM, which covered from 317
46,189,407-56,107,784 bp on the physical map (Figure 4a, Table 2).
318
On LG6, QTLs were only detected on the male map for three years for PM and for two 319
years (2017 and 2018) for PF (Figure 2 and Table 2). For PM (2016, 2017 and 2018) 320
and PF (2017), the peak marker is the same, Rh12GR_56601_1304 (29.7 cM, located 321
on chr6 at 31,814,891 bp). For PF2018, the peak marker is Rh88_37299_454 (11.5 cM, 322
located on chr6 at 5,410,244 bp). These QTLs explained between 5.28 and 8.45% of 323
the observed variance. The common 0.95 credible interval was from 15.59 to 42.49 cM, 324
which covered from 8,578,645 to 44,264,630 bp on the physical map (Figure 4b, Table 325
2).
326
On LG1, QTLs were only detected on the male map for PF for two years (2016 and 327
2018), and explained 6.52 and 6.99% of phenotypic variance, respectively. The 328
common 0.95 credible interval was at 12.78-44.11 cM, which covered from 329
20,231,658-62,553,371 bp on the physical map (Figure 4b, Table 2).
330
16 We checked the interaction between OB3@Rh12GR_16570_782 and OB4@Rh12GR_
331
60129_183, and between RW3@Rh12GR_52506_1218 and
332
RW6@Rh12GR_56601_1304, and no significant interaction was detected.
333
2.2 Two-part QTL analysis 334
In order to extend the analysis even further, we performed a two-part QTL analysis to 335
test the penetrance (presence/absence of prickles, LOD(p) were calculated with binary 336
traits) and the severity (density of prickles on stems with prickles, LOD(µ) were 337
calculated with non-zero quantitative phenotype) of these QTLs.
338
For the hypothesis LOD(p) on the female and male maps, we obtained a significant 339
LOD(p) on the LG3 for the two types of stem (PF and PM) and the three years (Figure 340
3, Supplementary Table 2). The marker with the highest LOD score on the OB map is 341
the same: Rw35C24 (SSR marker) located at 44.4cM (Chr03: 40,215,502 bp). This 342
QTL explained 13.38% to 16.72% of the variation. The peak marker on the RW map is 343
also the same for PF and PM for the three years: Rh12GR_52506_1218 located at 42.6 344
cM (42,317,122 bp). This QTL explained 20.69 to 33.21% of the variation. These data 345
suggested that the QTL detected on LG3 mainly controls the presence/absence of 346
prickles. Moreover, the LOD(p) on LG2 and LG6 for the male map were only 347
significant in PF2016 and PM2016, respectively (Figure 3), and they showed a weak 348
effect with an explanation of 1.80% and 2.70% of the variance, respectively 349
(Supplementary Table 2).
350
For the LOD(µ) hypothesis, we detected a significant QTL on the female map on LG4 351
for PM (2016 and 2017) and PF (2016) (Figure 3, Supplementary Table 2). The QTLs 352
17 explained 9.02% to 9.88% of the observed phenotypic variances. Therefore, this QTL 353
might be involved in the control of prickle density. On LG3, a significant QTL was 354
detected on the male map for PM (2016, 2017, 2018) and PF (2016), suggesting that a 355
QTL on LG3 might also control prickle density. This QTL is in the same region of the 356
QTL detected for penetrance (Figure 3, Supplementary Table 2). On LG1, the LOD(µ) 357
peaks in OB (PM2018) and in RW (PF2018) were higher than the genome-wild 358
threshold (µ); these QTLs explained 6.66% and 7.80%, respectively, of prickle density 359
variation.
360
2.3- The interaction of the LG3-QTL allele between OB and RW 361
Based on non-parametric and two-part methods, we identified QTLs for the presence 362
of prickles on LG3 for the OB and RW maps in the same region. To further investigate 363
how the alleles on these QTLs affect the presence of prickles, we visualized the number 364
of prickles for each genotype in the hybrid population depending of the Mendelian 365
distribution of the SNP markers at the LOD peak (Figure 5). The female and male 366
alleles are referred to as a,b and c,d, respectively. The separation ratio ac:ad:bc:cd in 367
offspring is 33:54:16:48, and was significantly different from the expected segregation 368
of 1:1:1:1 (37.5 for each) with a p-value = 0.004 estimated by a chi-squared test (Figure 369
5).
370
For PF and PM in all three years, we clearly see that the bd allele combination in hybrids 371
is correlated with no-prickle individuals or individuals with only a few prickles (less 372
than two on four internodes), whereas ac, ad and bc genotypes present prickles (Figure 373
5). These results suggest a dominant/recessive model for this QTL with the b and d 374
18 alleles linked to the null or recessive alleles (prickless mutant) and the a and c alleles 375
linked to the dominant alleles (prickles). For PM, a co-dominant effect can be detected 376
since the phenotype for ac is significantly different from the one for ad and bc (ac > ad 377
and ac > bc, p-value < 0.05, except for PM2016 between ac and ab; Figure 5), even if 378
the effect is weak (no large difference between the mean for ac and ad/bc). For PF, no 379
co-dominant effect was detected.
380
We also observed some odd phenotypes. For instance, OW9067 and OW9068 (red dots) 381
had no prickles and were grouped in the ad genotype, perhaps due to recombination 382
between the marker and the prickle locus (Figure 5). For individuals with the bd 383
genotype, six individuals (blue and green dots) always have prickles: OW9062, 384
OW9021, OW9052 and OW9109 (blue dots) look like the usual prickle genotypes and 385
are probably caused by recombination, but the two extreme exceptions, OW9106 and 386
OW9107 (green dots) with the highest prickle density are not that easy to clarify.
387
Moreover, some individuals exist with both prickly and glabrous stems in the same 388
plant.
389
3- Candidate genes in the QTL interval region and gene expression analysis 390
3.1 Candidate gene characterization and location in rose.
391
Since it was proposed that prickles originate from a deformation of glandular trichomes 392
in rose (Kellogg et al. 2011), we looked for rose homologues of transcription factors 393
(TF) known to be involved in the molecular control of trichome initiation and 394
development in Arabidopsis. The information from 15 TFs such as the bHLH (basic 395
helix-loop-helix), C2H2 Zinc-Finger, MYB, WD40 repeat and WRKY families are 396
19 presented in Supplementary Table 3. For a more detailed annotation, we performed 397
phylogenetic analyses on these protein families (Supplementary Figure 3).
398
Concerning the bHLH family (Supplementary Figure 3a), RC7G0190300, 399
RC1G0342400 and RC6G0407800 showed strong similarity with GLABROUS3, 400
MYC1 and TT8, respectively, where all of the proteins are in the same clade. They are 401
referred to as RcGL3, RcMYC1 and RcTT8, respectively.
402
For the C2H2 family, RC3G0150000, RC4G0390900 and RC4G0476500 are closely 403
related to GLABROUS INFLORESCENCE STEMS proteins (GIS, GIS2 and GIS3) 404
and ZINC FINGER PROTEIN (ZFP5, 6 and 8). RC3G0150000 seems to be more 405
closely related to GIS2, RC4G0390900 to GIS3 and RC4G0476500 to ZFP5.
406
RC2G0415300 and RC6G0454700 are related to ZFP1 and ZFP3 and AT5G10970.
407
They are referred to as RcZFP1-like1 and RcZFP1-like2, considering that they are 408
closer to ZFP1 (Supplementary Figure 3b).
409
R2R3 MYB and R3 MYB belong to the MYB family (Supplementary Figure 3c). In 410
the R2R3 MYB sub-family (blue sub-tree), RC7G0156100 is in the same clade as 411
GLABROUS1, whereas RC2G0033100 and RC7G0261400 are more closely related to 412
MYB82 and TT2, respectively. RC3G0322900 is in the same clade as MYB61, MYB50 413
and MYB86. In the R3 MYB sub-family (red sub-tree), RC2G0548400, RC1G0560100 414
and Chr1g0359121 (Raymond et al. 2018) are in the same clade of CPC, TRY, ETC1 415
and ETC3. RC1G0560100 and Chr1g0359121 are more closely related to TRY and 416
CPC, and are referred to as RcTRY and RcCPC, respectively. (Supplementary Figure 417
3d).
418
20 In the WD40 family, RC1G0586100 showed a strong similarity to TRANSPARENT 419
TESTA GLABRA 1(TTG1), and RC3G0186600 and RC2G0693200 also belong to this 420
clade.
421
In the WRKY family, as previously shown by Hibrand-Saint Oyant et al. (2018), 422
RC3G0244800 shows a strong similarity with AtTTG2 (TESTATRANSPARENT 423
GLABRA2), whereas RC3G0309600 and RC3G0309700 seem to be more closely 424
related to WRKY54 and WRKY70, RC3G0392200 to WRKY74, and RC3G0414600 425
appears to be related to WRKY34 and WRKY2.
426
We then located these rose homologue genes on the rose genome and looked for co- 427
location between these genes and the QTLs previously described (Figure 4A and B).
428
Concerning the QTLs on LG3 (male and female map, Figure 4a and b), the most 429
interesting TF among the detected genes was RcMYB61 (RC3G0322900, at Chr03:
430
39,896,892-39,899,077) located in the cQTL interval (36.517-46.440 Mb) for the 431
female map (Figure 4A). As previously described (Hibrand-Saint Oyant et al. 2018), a 432
homologue of TTG2, a WRKY transcription factor (RC3G0244800), is also located in 433
the credible interval. RcGIS2 (RC3G015000), a GIS2 homologue is also located on LG3 434
but not in the cQTL interval. In addition to the candidate TFs, we also scanned the other 435
TFs co-located in the cQTL interval on LG3 of the female map. There are four bHLH, 436
two C2H2, three R2R3MYB and seven WD40 transcription factors (Supplementary 437
Figure 3, in blue) located under the cQTL.
438
Concerning the cQTL interval on LG4, RcGIS3 is positioned at Chr04: 50,315,805 - 439
50,317,009 (1.21 Kb), and near the peak marker Rh12GR_55601_ 1304 (52,239,028 440
21 kb) on the female map (Figure 4A). RC4G0476500, a ZFP5 homologue, is also located 441
on the female LG4 but not in the QTL interval.
442
Concerning the QTL on the male LG1, RcMYC1, RcTRY and RcCPC, which are 443
positioned at 44,468,298-44,473,643 bp, 47,708,966-47,709,896 bp and 62,070,383- 444
62,072,848 bp, respectively, are located in the cQTL region (20.232Mb-62.553Mb) of 445
PF (2016, 2018) on the male LG1. The gene RC1G0586100 (RcTTG1) is also located 446
on LG1 but outside this interval.
447
For the male LG6, RC6G0407800, a homologue of TT8, is not located in the cQTL 448
credible interval, and no studied gene was detected below this QTL.
449
3.2 Candidate gene expression in glabrous and prickle roses.
450
Based on the positional approach, we identified ten interesting candidate genes, six 451
within the QTL interval and the other four outside of QTL but near the credible interval 452
(Figure 4). In order to obtain more information about these genes, we studied their 453
transcript accumulation by RT-qPCR in tissues from prickle (P) and prickless (NP) 454
stems at different developmental stages: I, IIa, IIb, IIc, III (Supplementary Figure 1).
455
The cluster analysis of gene expression clearly showed that all the samples can be 456
divided into two main groups: PI, NPI, PIIa, NPIIa, PIIb, NPIIb were gathered into one 457
group, and PIIc, NPIIc, PIII, NPIII into another group (Figure 6a). At the sup-group 458
level, PI and NPI, PIIa and NPIIa, PIIb and NPIIb, PIIc and NPIIc were clustered 459
together, respectively. At the same stem developmental stage, prickle and glabrous 460
samples (P and NP) behave similarly, suggesting no major difference of transcript 461
accumulation between prickle and glabrous samples; the observed differences are more 462
22 closely related to stem development.
463
To extend the analysis even further, we used NP as a control to compare the normalized 464
expression of genes between P and NP samples in the different stages (Figure 6b). In 465
stage I, two genes are differentially expressed: RcMYB61 and RcGIS2 were down- 466
regulated in prickly stems, with a significant p-value = 4.1e-5 and 2.9e-4 (Figure 6b), 467
respectively. In stage IIa, only RcZFP5 was significantly differentially expressed 468
between P and NP, with a p-value = 0.0056 and FC = -5.7606 (Figure 6b). A different 469
pattern is observed in stage IIb where RcZFP5 expression was up-regulated with FC = 470
8.2240 and a p-value = 0.0025. In addition, the transcripts of RcMYC1, RcCPC and 471
RcGIS2 were also significantly accumulated (p-value = 4.1e-5, 0.0048, 0.0012, 472
respectively) in stage IIb. In stage IIc, no significant change in gene expression was 473
detected. In stage III, the RcGIS2 transcript is differentially accumulated with FC = - 474
4.908 and a p-value = 0.043. The same pattern is observed for RcMYB61 with a p- 475
value = 4.9e-4.
476
We followed the transcript accumulation during stem development in prickly and 477
glabrous stems (NPI as a control; Figure 6c). All the studied genes are regulated 478
between the different samples. For instance, RcMYB61 is up-regulated and RcMYC1 is 479
down-regulated between the different stages. For RcZFP5, we observed a delay in the 480
decrease of transcript accumulation, with a decrease in stage IIa for glabrous stems and 481
in stage IIb for stems with prickles (Figure 6c).
482
Discussion
483
Two types of prickles are present in the OW progeny, originating from different 484
23 structures.
485
A good understanding of prickle morphology is required to serve as the foundation for 486
genetic and molecular studies. We identified two different types of prickles in our 487
population: it appears that GP and NGP originate from glandular and non-glandular 488
structures, respectively. This conclusion is different from previous studies in rose, 489
which reported that prickles were extensions or modifications of glandular trichomes 490
(Kellogg et al. 2011), and in other species (Ma et al. 2016b; Pandey et al. 2018). Asano 491
et al. (2008) observed two types of prickles in the cultivated rose ‘Laura’, described as 492
large size and small size prickles. The large size prickles look similar to NGPs in our 493
study. The small size prickles, referred to as acicles (Asano et al. 2008), are more 494
closely related to the glandular prickles (GP) we observed since they have a glandular 495
head that accompanies them throughout their lifetime. The difference between these 496
two types of prickles is also related to their segregation in the OW population (Figure 497
1d), demonstrating that different genetic determinisms are involved. In this study, since 498
only a few F1 individuals had GPs, we cannot perform a genetic analysis on GPs, we 499
concentrated our analysis on NGPs.
500
A complex genetic determinism for prickles in rose 501
Prickles on stems exhibited transgressive segregation in diploid OW, the same as for 502
the tetraploid K5 population (Koning-Boucoiran et al. 2012; Gitonga et al. 2014;
503
Bourke et al. 2018), supporting the hypothesis that multiple loci may be responsible for 504
this trait.
505
24 Using the ‘non- parametric’ QTL approach, we detected a stable QTL on LG3 in the 506
three different years for both types of stems (PM and PF) on both the male and female 507
genetic maps. We also demonstrated that this QTL mainly controls the 508
presence/absence of prickles (Figure 3) using the ‘two-part’ QTL method. Interestingly, 509
for PM in males, the QTL on LG3 may also be involved in regulating prickle density 510
(severity in the two-part QTL analysis; Figure 3). A similar phenomenon was observed 511
for the petal number with a locus on LG3 that controls the difference between simple 512
and double petals, and a variance of the petal number that exists within the double petal 513
flower is controlled by another locus (Roman et al. 2015; Hibrand-Saint Oyant et al.
514
2018).
515
We further enhanced the description of QTLs on LG3 that affect the presence/absence 516
of prickles. A significantly distorted segregation was observed at the peak marker 517
position. That unusual segregation ratio might be explained by the presence of a self- 518
incompatibility locus (Hibrand-Saint Oyant et al. 2018) near the peak marker for this 519
QTL. On the basis of the phenotype-genotype relationship (Figure 5), we proposed that 520
the PRICKLE alleles on this QTL are both heterozygous (np/P) in OB and RW, and that 521
the presence of prickles is controlled by a dominant allele (np/P or P/P), and that the 522
glabrous stem in the progeny is due to the combination of the two recessive alleles 523
coming from both parents (np/np). These results are important for breeders who need 524
to combine recessive alleles to obtain glabrous roses, an allelic combination that can be 525
difficult in tetraploid roses. Development of specific molecular markers of the recessive 526
allele may by useful for breeders. However, it should be noted that the actual markers 527
25 used (peak of the QTL) are only closely linked to the PRICKLE locus and few 528
recombinants are observed in the progeny. Furthermore, the phenotype of the 529
individuals with the two recessive alleles (bd phenotype; Figure 5) are not stable and 530
some of the hybrids were regularly seen to develop some prickles on parts of the stems.
531
Indeed, this phenomenon is widespread in roses. Rose breeders have reported that 532
glabrous mutants have either been unstable for the prickless trait (Nobbs 1984; Rosu et 533
al. 1995), or reverted to the prickly character after a freezing winter or other 534
environmental stresses (Nobbs 1984; Oliver 1986; Druitt and Shoup 1991; Canli 2003).
535
Taken together, we assumed that a single major locus on LG3 controlled the 536
absence/presence of stem prickles. Further investigations are necessary to more closely 537
identify molecular markers (for molecular assisted breeding) and the mechanisms 538
behind the instability.
539
In the ac, ad and bc genotypes, each genotype has a continuous quantitative trait, 540
indicating that there are other loci responsible for prickle density variance. Other QTLs 541
affecting quantitative traits were detected on LG4 in OB and on LG1 and 3 in RW (Two- 542
part QTL analysis; Figure 3). The LG4 QTL has a strong effect on PM but a weak effect 543
on PF. For the QTL on LG1, it only had a weak effect on PF and on PM in 2018. Those 544
three loci are related to the density of prickles, indicating that there are multiple genes 545
responsible for the density trait, and that those genes have a different effect on the 546
different stems.
547
Detected QTLs are conserved in the Rosa genus and the Rosideae subfamily 548
Thanks to the link between genetic maps and reference genome sequences (Hibrand- 549
26 Saint Oyant et al. 2018), we were able to compare our results with previous genetic 550
studies by associating genetic map markers.
551
A QTL was previously detected on LG3 in different diploid and tetraploid populations 552
(Crespel et al. 2002; Linde et al. 2006; Koning-Boucoiran et al. 2012; Hibrand-Saint 553
Oyant et al. 2018 ; Bourke et al. 2018), which is consistent with our results: a strong 554
QTL on Chr3 with a high LOD value was detected in all of the environments (across 555
and between years and types of stems). This demonstrated that Chr3 QTL is a robust 556
QTL detected independently of ploidy and the environment, and is present in various 557
genetic backgrounds.
558
Recently, three QTLs on LG3, 4 and 6 were detected in the tetraploid K5 population 559
with a high density of SNPs genetic map (Bourke et al. 2018). Interestingly, the QTLs 560
identified from the diploid (OW) were almost identical to tetraploid (K5) populations 561
(LG3, 4 and 6), with the slight difference that we also detected a weak QTL on LG1, 562
which was only significant in males for two of the years. This slight difference might 563
be due to the genetic background of the parents of the K5 and OW populations. In 564
fact, in K5 populations, one parent is prickly and the other glabrous, whereas in OW 565
populations, both parents have prickles. Bourke et al. (2018) reported that two SNP 566
markers, K7826_576 (located on the Chr3: 37,706,920 pb) and K5629_995 (located on 567
the Chr4: 57,791,999 bp) are linked to the stem prickle trait. When compared with our 568
results, K7826_576 is located within our Chr3 cQTL interval region (36,517,224- 569
46,440,369 bp; Figure 4), and K5629_995 is very close to our Chr4 QTL interval 570
(46,189,407-56,107,784 bp). These results suggest that QTLs detected on LG3 and 4 571
27 could be similar between OW and K5 progenies.
572
In Rosaceae, the genetic determinism of prickle was studied in raspberry (Rubus 573
idaeus), where two QTLs were detected on LG4 and 6 (Molina-Bravo et al. 2014).
574
Using synteny viewer tools (https://www.rosaceae.org/synview/search; Jung et al.
575
2014), we checked the synteny. The region where the QTL is located on LG6 in R.
576
occidentalis (position 6,028Mb) is syntenic with a region on rose chromosome 2 577
(position 42,330 Mb), where no QTL for prickle density was detected in our study. The 578
region where the QTL 4 is located (position 0.101 Mb) is syntenic with the region on 579
rose chromosome 4 (position 58,768 Mb), very close to the main QTL we detected on 580
this chromosome (Table 2). These results could suggest that the two QTLs in rose and 581
raspberry might be syntenic and share a common evolutionary history. In another 582
publication, Graham et al. (2006) identified the gene H that controls cane pubescence.
583
The locus is mapped on LG2, which is syntenic with the rose LG6 where one of the 584
QTLs is located, detected in R. x wichurana. However, no precise location is available 585
to allow us to assume a possible common origin.
586
Candidate gene below the QTL interval region 587
Prickles are assumed to originate from a ‘trichome-like structure’. In order to find a 588
putative candidate gene for the identified QTLs, we looked for homologue genes known 589
to be involved in trichome initiation and development in Arabidopsis. We annotated 15 590
rose TFs that, based on similarity, can be involved in trichome development in rose:
591
RcGL1, RcMYB82, RcMYB61, RcCPC, RcTRY, RcGL3, RcTT8, RcMYC1, RcTTG1, 592
RcTTG2, RcZFP5, RcGIS3, RcGIS2, RcZFP1 and RcGL2 (Table 3). Among them, a 593
28 few were below the detected QTLs: RcMYB61 and RcTTG2 below the QTL on LG3;
594
RcGIS3 below the QTL on LG4; and RcCPC, RcTRY and RcMYC1 below the QTL on 595
LG1. ZFP5 (Chr04: 57,125,905 bp) is out of the QTL interval on LG4 in OW, but close 596
to the peak LOD marker K5629_995 of QTL in the K5 population (Chr04: 57,791,999 597
bp) (Bourke et al. 2018). These genes are good candidates for the detected QTLs.
598
Candidate genes transcript expression in glabrous and prickle F1 individuals 599
We quantified ten TF gene transcripts in glabrous and prickle F1 individuals in different 600
developmental stages using RT-qPCR. Surprisingly, minor differences were observed 601
between glabrous and prickle samples, with the main differences occurring between 602
developmental stages (as demonstrated by the heatmap analysis, Figure 6a). Based on 603
transcript accumulation, this suggests that these homologues, known to be involved in 604
trichome initiation and development in Arabidopsis, are not implicated in prickle 605
initiation in rose, leading to the hypothesis that the two processes (trichome initiation 606
and prickle initiation) might involve different gene pathways. The candidate gene 607
approach may not be appropriate and a non-a priori approach such as a transcriptomic 608
analysis could be done between individuals with and without prickles.
609
Nevertheless, some differences in transcript accumulation are observed between 610
candidate genes. In the early stage (stage I), only RcMYB61 and RcGIS2 are slightly 611
more highly accumulated in glabrous stems. However, GIS2 and MYB61 are positive 612
regulators of trichome initiation (Gan et al. 2006), which is difficult to reconcile with 613
an increase in transcript accumulation in glabrous stems (Figure 6). Negative feedback 614
29 regulation during prickle initiation can explain this point, as regularly observed in 615
trichome initiation (Pattanaik et al. 2014) or, perhaps, differences are not at the 616
transcriptional level. It could be interesting to sequence the genes in the two parents to 617
see if a mutation can explain the phenotype.
618
RcZFP5 may also be an interesting candidate gene. This gene showed a different 619
regulation between glabrous and prickly stems. At stage IIa, RcZFP5 shows a strong 620
down-regulation in glabrous tissue, whereas this down-regulation is observed later at 621
stage IIc in tissues with prickles (Figure 6C). Furthermore, this gene is close to the QTL 622
on LG4. Its early repression in glabrous stems might explain why no prickles developed.
623
In A. thaliana, ZFP5 controls trichome initiation through GA signaling (Zhou et al.
624
2011). These data (concerning ZFP5 and MYB61) might suggest an implication of GA 625
in prickle development. However, this hypothesis needs to be functionally validated in 626
rose.
627
Conclusion
628
Prickle structure is an undesirable trait, not only in rose but in most crops in general.
629
We identified a complex genetic determinism with a major locus on LG3 that controls 630
the presence of prickles and a few QTLs that control prickle density. Further studies are 631
necessary to develop markers for breeding selection and to identify the molecular bases.
632
Using a candidate gene approach, we proposed different hypotheses concerning the 633
gene involved in prickle initiation in rose. Approaches such as transcriptomics may help 634
to identify new key regulators of prickle initiation and development in rose.
635
30
Acknowledgements
636
We are grateful to the experimental unit (UE Horti) for their technical assistance in 637
plant management, and the ImHorPhen team (D. Besnard, R. Gardet) of IRHS for 638
taking care of the plant cuttings in the greenhouse. We would also like to thank the 639
IMAC technical platforms (F. Simonneau, A. Rolland) of SFR Quasav for supervising 640
the histological experiment, and the PTM ANAN (M. Bahut) of the SFR Quasav for 641
overseeing the RT-qPCR experiment. We acknowledge J. Chameau of the GDO team 642
for helping to obtain the different stages of the sample.
643
This work was supported by funding from the National Natural Science Foundation of 644
China (31760585), the China Scholarship Council ([2017]3109) and the Natural 645
Science Foundation of Yunnan (2016FB061).
646
Figure legends
647
Figure 1 Different types of prickles on the OW progeny stem and their distribution. (a) 648
Stem prickles in the female ‘Old Blush’ (OB) and the male R. x wichurana (RW); NGPs:
649
non-glandular prickles. (b) Stem prickles in F1 progeny. Glabrous: no prickles 650
whatsoever on the recorded stems in the three years. (c) Macroscopic photos of the 651
terminal part of the stems with different types of prickles (number of offspring); GPs:
652
glandular prickles. (d) The distribution and Q-Q plot of NGPs and GPs in the F1
653
progeny in 2018; PF: prickles on the floral stem; PM: prickles on the main stem.
654
Figure 2 LOD curves of the QTL scan for the NGPs on the floral (FM) and main (PM) 655
stems in (a) female (OB) and (b) male map (RW) calculated with a non-parametric 656
model for the three years (2016, 2017 and 2018, with red, blue and green lines, 657